I've got a kernel that produces decent results on my laptop, but I can't get it to run properly on Kaggle. With GPU enabled, it fails with a memory error. Without GPU, it times out. I'm trying to perform LSI using Gensim. Does anyone have any suggestions for how to solve this?
", "messages": [ { "id": "424115", "postDate": "11/19/2018 15:37:22", "content": "I've got a kernel that produces decent results on my laptop, but I can't get it to run properly on Kaggle. With GPU enabled, it fails with a memory error. Without GPU, it times out. I'm trying to perform LSI using Gensim. Does anyone have any suggestions for how to solve this?
", "rawMarkdown": "I've got a kernel that produces decent results on my laptop, but I can't get it to run properly on Kaggle. With GPU enabled, it fails with a memory error. Without GPU, it times out. I'm trying to perform LSI using Gensim. Does anyone have any suggestions for how to solve this?", "votes": null }, { "id": "424142", "postDate": "11/19/2018 16:26:55", "content": "Yeah recently a lot of people I've talked to have been having issues with Kaggle kernels. There are issues using them in general (execution errors, memory errors, kernels deleting saved work, etc.) and I've heard a few specific ones relating to use of GPU (which seems to be broken ATM). Kaggle's provision of a free cloud kernel is a nice sentiment but it's execution has been filled with errors. I haven't heard of anyone who figured out a way around the GPU issue and based on what I've heard Kaggle support will do for you (nothing) I'd recommend just continuing to run your code on your laptop or using Google Collab and then submitting manually.
", "rawMarkdown": "Yeah recently a lot of people I've talked to have been having issues with Kaggle kernels. There are issues using them in general (execution errors, memory errors, kernels deleting saved work, etc.) and I've heard a few specific ones relating to use of GPU (which seems to be broken ATM). Kaggle's provision of a free cloud kernel is a nice sentiment but it's execution has been filled with errors. I haven't heard of anyone who figured out a way around the GPU issue and based on what I've heard Kaggle support will do for you (nothing) I'd recommend just continuing to run your code on your laptop or using Google Collab and then submitting manually.", "votes": null }, { "id": "424599", "postDate": "11/20/2018 11:50:41", "content": "The trouble is, this is a Kernels-only competition, so I can't submit manually.
", "rawMarkdown": "The trouble is, this is a Kernels-only competition, so I can't submit manually.", "votes": null }, { "id": "424625", "postDate": "11/20/2018 12:54:32", "content": "The GPU has a certain memory limit (I think 12GB) and of course a 2 hour runtime limit. This forces you most likely to stick to simpler models and algorithms. You should compare the specs of the CPU and GPU runtimes with your consumption on the laptopt to see what the issue might be.
", "rawMarkdown": "The GPU has a certain memory limit (I think 12GB) and of course a 2 hour runtime limit. This forces you most likely to stick to simpler models and algorithms. You should compare the specs of the CPU and GPU runtimes with your consumption on the laptopt to see what the issue might be.", "votes": null }, { "id": "424628", "postDate": "11/20/2018 12:55:16", "content": "You cannot manually submit in this competition.
", "rawMarkdown": "You cannot manually submit in this competition.", "votes": null } ], "comments": [ { "id": 424142, "author_name": "alecthekulak", "author_url": "", "post_date": "11/19/2018 16:26:55", "content": "Yeah recently a lot of people I've talked to have been having issues with Kaggle kernels. There are issues using them in general (execution errors, memory errors, kernels deleting saved work, etc.) and I've heard a few specific ones relating to use of GPU (which seems to be broken ATM). Kaggle's provision of a free cloud kernel is a nice sentiment but it's execution has been filled with errors. I haven't heard of anyone who figured out a way around the GPU issue and based on what I've heard Kaggle support will do for you (nothing) I'd recommend just continuing to run your code on your laptop or using Google Collab and then submitting manually.
", "votes": null, "replies": [ { "id": 424628, "author_name": "philippsinger", "author_url": "", "post_date": "11/20/2018 12:55:16", "content": "You cannot manually submit in this competition.
", "votes": null, "replies": [] } ] }, { "id": 424599, "author_name": "petebleackley", "author_url": "", "post_date": "11/20/2018 11:50:41", "content": "The trouble is, this is a Kernels-only competition, so I can't submit manually.
", "votes": null, "replies": [] }, { "id": 424625, "author_name": "philippsinger", "author_url": "", "post_date": "11/20/2018 12:54:32", "content": "The GPU has a certain memory limit (I think 12GB) and of course a 2 hour runtime limit. This forces you most likely to stick to simpler models and algorithms. You should compare the specs of the CPU and GPU runtimes with your consumption on the laptopt to see what the issue might be.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "424115": "I've got a kernel that produces decent results on my laptop, but I can't get it to run properly on Kaggle. With GPU enabled, it fails with a memory error. Without GPU, it times out. I'm trying to perform LSI using Gensim. Does anyone have any suggestions for how to solve this?", "424142": "Yeah recently a lot of people I've talked to have been having issues with Kaggle kernels. There are issues using them in general (execution errors, memory errors, kernels deleting saved work, etc.) and I've heard a few specific ones relating to use of GPU (which seems to be broken ATM). Kaggle's provision of a free cloud kernel is a nice sentiment but it's execution has been filled with errors. I haven't heard of anyone who figured out a way around the GPU issue and based on what I've heard Kaggle support will do for you (nothing) I'd recommend just continuing to run your code on your laptop or using Google Collab and then submitting manually.", "424599": "The trouble is, this is a Kernels-only competition, so I can't submit manually.", "424625": "The GPU has a certain memory limit (I think 12GB) and of course a 2 hour runtime limit. This forces you most likely to stick to simpler models and algorithms. You should compare the specs of the CPU and GPU runtimes with your consumption on the laptopt to see what the issue might be.", "424628": "You cannot manually submit in this competition." }, "source": "meta" }